Empirical quantitative case study in operations management

Name: Seongmin Moon

University: Newcastle University Business School

Address: Citywall, Citygate, St James Boulevard, Newcastle Upon Tyne, NE1 4JH,

United Kingdom

Telephone: (+44) 7796785303

Email: seongmin.moon@ncl.ac.uk

Supervisor, Name: Dr Andrew Simpson

Supervisor, Email: andrew.simpson@ncl.ac.uk

16TH EDAMBA Summer Academy

Soreze, France

July 2007

1
Abstract

This paper establishes the methodological framework for my research. My research can be

classified as empirical normative quantitative research and empirical quantitative case study

in operations management discipline. This is also can be considered as exploratory research

and explanatory research.

Single case research design was employed for my research, because the single case design

provides more opportunities for in-depth observation, and the single case design is

appropriate for the extreme case, the typical case and the longitudinal case which are the

cases of this study.

Keywords: operations management, case study, theory

2
This paper delineates the research methodology for my research. It starts by identifying the

purpose of this research. In section 2, the methodological categories in operational

management research are reviewed. In section 3, the suitability of the case study to research

question is discussed. In section 4, the relation of theory to the case study is examined. In

section 5, the case study research design is outlined. In section 6, the criteria for evaluating

the research design are discussed. In section 7, the research procedure of this study is

described. Finally, this paper is summarised in section 8.

1. Purpose of research

The purposes of research can be categorised as exploratory, descriptive and explanatory

(Saunders et al., 2007). An exploratory study can be described as finding out what is

happening, and asking questions and assessing phenomena in a new light; a descriptive study

can be described as portraying an accurate profile of persons, events or situations; and an

explanatory study can be described as establishing causal relationships between variables.

Recalling the aim of this research, the aim is “to identify an appropriate forecasting system

for the spare parts demand in the SKN”. The aim is achieved by answering research questions,

which require ascertaining the nature of demand and the nature of the forecasting system in

the SKN in the new forecasting strategy, as well as establishing the relationships between

demand features and forecasting performances. Therefore, this study appears to be in line

with an exploratory study as well as an explanatory study.

2. Empirical research in operations management

Operations management is defined as “the effective planning, organising, and control of all

3
resources and activities necessary to provide the market with tangible goods and services. It

applies to manufacturing, service industries and not-for-profit organisations” (Waller, 2003, p.

875). This study focuses on model-based quantitative research in operations management

discipline applied to a not-for profit organisation (i.e. the SKN). Model-based quantitative

research refers to the research where models of causal relationships between control variables

and performance variables is developed, analysed or tested (Bertrand and Fransoo, 2002). In

causal relationships, a change of value α in one variable will lead to a change of f(α) in

another variable, so that a model can be utilised to predict the future state of the modelled

processes (Bertrand and Fransoo, 2002). In this study, future state refers to the change of

performance variables such as inventory level and total costs. The change of forecasting

method for spare parts demands can lead to a change in the performance variables.

Bertrand and Fransoo (2002) classified quantitative operations management research as

axiomatic research and empirical research: a) axiomatic quantitative research indicates the

process of achieving resolutions by the defined model; b) empirical quantitative research

indicates the process of achieving resolutions by empirical findings. Bertrand and Fransoo

(2002) divided axiomatic and empirical quantitative research into two sub-categories, namely

normative research and descriptive, by their objectives: normative research has the objective

of establishing policies, strategies and actions; descriptive research has objective of analysing

model or describing the causal relationships in reality.

In terms of the classification scheme from Bertrand and Fransoo (2002), this research might

be in line with empirical normative quantitative research, because the resolutions were

achieved by empirical findings; and the objective of this study is to establish a forecasting

policy for the spare parts by collating competing forecasting strategies (i.e. DF, TD, CF). EN

4
was claimed difficult to be verified, because controlling all relevant variables is impossible

which is required for evaluating performance changes in EN (Bertrand and Fransoo, 2002).

Verification is an important issue for this study. A simulation experiment was employed to

evaluate the performance of forecasting methods in the controlled environment similar to

reality.

Wacker (1998) classified operations management research as analytical (formal) research and

empirical research for the purpose of theory building (see Table 1). In empirical statistical

research, theoretical relationships are verified statistically in large external samples from

reality; however, an empirical case study looks into small samples to test and develop

complex relationships between variables to suggest a new theory (Wacker, 1998).

In terms of the classification scheme from Wacker (1998), this research might be in line with

an empirical quantitative case study, because this study investigates only one case (i.e. the

SKN) to test and develop forecasting strategy for the spare parts demand.

5
 Table 1 Research category in operations management (Wacker, 1998)

Types of research included

Analytical Conceptual Futures research scenarios, introspective reflection,

hermeneutics, conceptual modelling
Mathematical Reason/logical theorem providing normative analytical
modelling, descriptive analytical modelling, proto-typing,
physical modelling, laboratory experiments, mathematical
simulation
Statistical Mathematical statistical modelling

Empirical Experimental Empirical experimental design, descriptive analytical

design modelling
Statistical Action research structured and unstructured research,
sampling surveying, historical analysis, expert panels
Case studies Field studies, case studies

3. Choice of case study strategy

A case study is defined as “an empirical inquiry that investigates a contemporary

phenomenon within its real-life context” (Yin, 2003, p. 13). A case study is a research

strategy which concentrates on perceiving the dynamics present within single settings

(Eisenhardt, 1989).

A case study is particularly good for examining “why” as well as “how” and “what”

questions (among question series: “who”, “what”, “where”, “how” and “why”), which are

enquiries about a contemporary set of events over which the investigator has little or no

control (Yin, 2003, Saunders et al., 2007). Especially, the “how” question is suitable for a

case study because this question deals with operational links needed to be traced over time,

rather than mere frequencies or incidence (Yin, 2003). Thus, the case study strategy is most

often employed in explanatory and exploratory research (Saunders et al., 2007).

6
Looking back at the research questions in this study, “how” and “what” questions can be

noticed (Table 2). These questions are appropriate to the case study. In addition, the case study

is suitable for exploratory research and explanatory research which are the cases of this study.

The research questions in this study lead both to theory testing and to theory development

(Voss et al., 2002).

Table 2 Research questions

a) What is the nature of the spare parts demand in the South Korean Navy (SKN)?

b) What forecasting method is appropriate for the spare parts demand in the SKN?

c) Under what conditions are TD or CF superior or inferior to DF?

d) How can the spare parts demand be classified in order to produce the most

superior forecast?

4. Theory in case study

A theory is defined as “a statement of relationships between units observed or approximated

in the empirical world” (Bacharach, 1989, p. 498). Every research is based on theory as

knowledge base.

4.1. Theory as knowledge base

There are three possibilities in which: a) existing theory provides no framework for findings;

b) existing theory conflicts with findings; and c) existing theory is in accordance with

findings (Eisenhardt, 1989, McCutcheon and Meredith, 1993, Yin, 2003). When the existing

theories are indigent and the available literature provides no conceptual framework or

hypotheses of note, such a knowledge base may not be a good theoretical foundation, and any

7
new empirical study is likely to assume the characteristic of an “exploratory” study (Yin,

2003).

On the other hand, the existing theories can be either in accord with the findings or in

disagreement with the findings; in both cases, theory is important (Eisenhardt, 1989).

Eisenhardt (1989) noted two reasons why enfolding literature which conflicts with the

emergent theory is important: first, if conflicting findings are ignored, then confidence in the

finding is reduced; second, conflicting literature generates an opportunity for refining the

theory.

Additionally, theories for a case study can be very well-developed, especially where the

object is to test or compare theories against empirical evidence, or the necessary foundation is

available in the well-developed theories from other disciplines (McCutcheon and Meredith,

1993). Such cases are also important, because these theories bind together underlying

similarities in phenomena typically not related to each other (Eisenhardt, 1989). This process

of linking results might be essential in theory-building case study research, because the

findings often rely upon a very restricted number of cases (Eisenhardt, 1989).

Theory is important in this study, in that the findings of this study are related to literature.

However, theories related to the hierarchical forecasting (HF) for non-normal demand such as

spare parts demand are not well-developed. Two research gaps were identified. However, the

necessary theoretical base is available from literature for HF strategy as well as DF strategy

in similar contexts.

8
4.2. Theory induction from data

Theory can be formed by either induction or deduction (Saunders et al., 2007). Wacker

(1998) pointed out that the pivotal distinction between a case study and an analytical method

is that empirical case study methods employ induction (i.e. depend on data) and the analytical

methods employ deduction. If a theory is based on data, then a large amount of data is

required, and case studies are a prime source of this research (McCutcheon and Meredith,

1993). The data can be quantitative data or qualitative data; they can be collected from either

single or multiple cases (Yin, 2003). This study adopts an empirical case study strategy which

employs an inductive method.

4.3. Generalisation

Eisenhard (1989) argued that binding the emergent theory with existing literature strengthens

the internal validity, generalisability (external validity), and theoretical level of theory

building from case study research. Internal validity demonstrates a causal relationship, in

which certain conditions lead to other conditions; and external validity tests whether a study’s

findings are generalisable beyond the immediate case study (Yin, 2003). Generalisability is a

particular concern for a sing case study as this research (Saunders et al., 2007). In this case,

analytic generalisation can be claimed for the case study research (Yin, 2003). There are two

kinds of generalisation from case to theory: statistical generalisation and analytic

generalisation (Yin, 2003).

In statistical generalisation, generalisability is established by an inference made about a

population on the basis of empirical data collected about a sample (Yin, 2003). However,

statistical generalisation should not be considered to be the method of generalising the results

9
of the case study (Yin, 2003).

In analytic generalisation, generalisability is established by the process as: an existing theory

is employed as a framework with which to collate the empirical results of the case study; then,

when more cases appear to support the same theory, replication can be claimed (McCutcheon

and Meredith, 1993, Yin, 2003). Analytic generalisation can be used in either single case or

multiple case study (Yin, 2003).

This study employs analytic generalisation in single case study design. A single case study

(i.e. the SKN) is used for advocating or refining existing theories. Then, the theory

established from the case study could extend to other situations such as other militaries and

business logistics.

5. Research design

Research design is the logical sequence that links the empirical data to a study’s initial

research questions; that is, the design discourages the situation in which the evidence is

disconnected from the initial research questions (Yin, 2003).

5.1 Four components of research design

Yin (2003, p. 21) identified five components of a case study research design: a) a study’s

questions; b) its proposition, if any; c) its unit(s) of analysis; the logic linking the data to the

propositions; and the criteria for interpreting the findings. Components a), b) and c) refer to

what data are to be collected, whereas components d) and e) refer to what is to be done after

the data have been collected (Yin, 2003). In this research, the four components a), c), d) and

e) are considered because this research dose not make research proposition.
10
The first component is research questions. Although a case study is an inductive approach, a

preliminary view of the general constructs or categories, and their relationships, is required;

then, initial research questions behind the proposed study should be followed (Voss et al.,

2002). Even though the prior questions are tentative, it is crucial to establish a well-defined

focus at the start, and to direct the collection of data (Voss et al., 2002). The research

questions of this research direct this research to focus upon the research topic, review the

related literature and collect the spare parts data in SKN.

The second component, unit of analysis is relevant to the fundamental problem of defining

what the ‘case’ is (Yin, 2003). Precisely specifying research questions leads into the

appropriate selection of the unit of analysis (Yin, 2003). As shown in Table 2, the research

questions seem to lead into the one unit of analysis; that is, the spare parts supply system in

the SKN. Once a general definition of the case has been established, other clarifications in the

unit of analysis become important: for example, specific group of people, district boundary,

or specific time boundary(Yin, 2003). Specific three kinds of warships are clarified in this

study. Since these three kinds of warships began to be commissioned in the early 1980s, they

have been utilised in major roles during Naval operations (Saunders, 2006). The time

boundary is decided from January 2002 to November 2007, because the Naval maintenance

data system, which is the major data source, has been stabilised since 2002.

The third component, linking data to propositions is a way of relating the data to the

propositions (Yin, 2003). In lieu of the proposition, research questions are considered to be

the objectives to link the data to. In order to relate the data to research questions a), the spare

parts demand and the current inventory control methods of the SKN in the SKN will be

11
analysed; to research question b), forecasts will be generated with the data by the competing

forecasting methods and compared with each other; to research question c), the superiority of

the competing forecasting methods under different condition such as the accuracy measures,

the equipment groups and the demand features will be examined; and to research question d),

the superiority of the competing forecasting methods in the classification model will be

examined.

The last component is the criteria for interpreting the findings. The experiment which adopts

competing forecasting methods is interpreted using an appropriate accuracy measure (i.e.

absolute measures and relative measures to another method). This will be verified by

simulation (i.e. derivative measure).

5.2 Choice of the single case design

Table 3 Choice of number of cases (Voss et al., 2002)

Choice Advantages Disadvantages
Single Greater depth Limits on the generalisability of conclusions
cases drawn. Biases such as misjudging the
representativeness of a single event and
exaggerating easily available data
Multiple Augment external validity Less depth per case
cases

Voss et al. (2002) discussed case study research design in terms of the number of cases. They

categorised it as single cases and multiple cases, and illustrated the advantages and

disadvantages as Table 3. Single cases have an advantage of greater depth. In this study, the

single case design (i.e. the case of the SKN) is expected to provide more opportunity for in-
12
depth observation.

However, this single case design might have limitations (Leonard-Barton, 1990, Voss et al.,

2002). Firstly, single cases have limits on the generalisability of the conclusions, because

models or theories are developed from one case study; secondly, the limit of generalisability

implies the risks of misjudging the representativeness of a single event, and of exaggerating

easily available data. These risks are also present in multiple cases, although these are

mitigated in multiple cases (Voss et al., 2002). However, as stated above, analytic

generalisation in lieu of statistical generalisation can be used for single cases as well as

multiple cases.

Yin (2003) postulated five rationales for single case designs (see Table 4). The case of spare

parts supply in the SKN might represent a typical military logistical case. This case can

represent an extreme case as well, for its extremely non-normal demand features. Moreover,

this case is a longitudinal case using data from 2002 to 2007. These rationales could serve as

the main reasons for conducting this single case study.

Table 4 Five rationales for single case design (Yin, 2003)

a) when it represents the critical case in testing a well-formulated theory;

b) when the case represents an extreme case or a unique case;

c) a single case is the representative or typical case;

d) a single case study is the revelatory case; and

e) a single case study is the longitudinal case: studying the same single case at two or

more different points in time.

13
Furthermore, Yin (2003) sub-categorised the categorisation as a two × two matrix: single-case

vs. multiple-case × holistic vs. embedded. A single case can involve more than one unit of

analysis: a case study design involving embedded units is called an embedded case study

design; a case study design examining only the global nature of an organisation is called a

holistic design (Yin, 2003). The case of spare parts supply in the SKN can be considered to

be a holistic single case design if the case is considered at the Naval supply centre of the

Naval Logistics Command (NLC); however, the case can be regarded as an embedded single

case design as well if the case is examined at the depots of the Naval bases. There are three

major Naval bases and five minor Naval bases in the SKN (Saunders et al., 2007). This study

employed the holistic single case design, because the focus of this study is to evaluate

forecasting performance rather than inventory system performance. Therefore, the complex

multi-echelon inventory system consisted of the eight depots (the first echelon) and the Naval

supply centre (the second echelon) in the SKN is not required to be described in detail. The

multi-echelon inventory system refers to the inventory systems in which materials flowing

through the system are stored at different points, before reaching the customer (i.e. warships)

(Slack et al., 2004).

6. Evaluating research design

Quantitative empirical case study research should be designed to test the validity of

quantitative theoretical models and quantitative theoretical problem solutions, with respect to

real-life operational situations (Bertrand and Fransoo, 2002). Yin (2003) discussed four tests

relevant to evaluating the quality of a research design: construct validity, internal validity,

external validity, and reliability.

Construct validity tests correct operational measures for the concepts being studied, and

14
ensure consistency between theory and the defined construct (McCutcheon and Meredith,

1993, Yin, 2003). Bertrand and Fransoo (2002) argued that operational research studies

generally lack construct validity, because data could be affected by subjective judgements.

However, the major data for this study, historical consumption of spare parts, was acquired

from the logistical database in the NLC. Therefore, construct validity appears not to be a

problem. However, the non-normal demand pattern causes distrust in the data generating

process.

Internal validity demonstrates a causal relationship, in which certain conditions lead to other

conditions (Yin, 2003). Internal validity is employed for explanatory (or causal) studies only

(Yin, 2003). This study is to investigate the causal relationships between forecasting methods

and forecasting performances. The causal relationships will be established by the various

accuracy measures; and validated by simulation. The internal validity of the classification

model will be established by the diagnostics of the model.

External validity tests whether a study’s findings are generalisable beyond the immediate case

study (Yin, 2003). External validity is a major barrier in conducting case studies (Yin, 2003).

As mentioned previously, a case study relies on analytic generalisation. In a single case study

design, the case of the SKN can be used in generalising existing theory. In addition, the case

of the SKN can also be used in extrapolating the theory to other situations (e.g. other

militaries and business logistics) relying on logical analysis.

Reliability tests whether the operations of a study can be repeated with the same results (Yin,

2003). Reliability was maintained by revealing every reference and every data source

explicitly, and presenting every equation and every process in models adopted in the study

15
transparently, so that any calculations are able to be audited. In order to maintain the

reliability of the forecasting performance, the forecasts generated were examined by a variety

of measures such as absolute, relative, and derivative measures.

7. Research procedure

The research procedure of this study is described as Figure 1. As a model-based single case

study, this study started from establishing the research questions in Chapter 1. As

aforementioned, research questions are crucial to establishing a well-defined focus at the start

and to direct the data collection. This is followed by theory development.

Figure 1Research procedure

In the theory development stage, the forecasting process of the SKN and its problems were

identified in Chapters 1 and 4. After reviewing literature relevant to these problems, the

research gaps are identified in Chapter 2. Then, initial study which combines initial data
16
selection, initial data collection, initial data analysis, and initial modelling is conducted. The

initial study results provide feedback which requires further literature review. This feedback

clarifies the research questions and guides further data collection. The stage of collecting data

is described in Chapter 4.

Following the collecting data stage, the modelling stage is carried out based on the theory

developed: the various competing forecasts are generated in Chapter 5; simulation is

conducted to verify forecasting performances in Chapter 6; and the classification model is

proposed in Chapter 7.

In the analysing and concluding stage, the results of empirical modelling is measured and

interpreted in view of forecasting accuracy as well as the inventory context in Chapters 5, 6

and 7. Thus, internal validity can be established by the various accuracy measures.

Depending on the results and the existing theory, the theory can be generalised as analytic

generalisation. Also, the theory is expected to contribute to supply policy development of the

SKN.

8. Summary

This paper presents a methodological framework for this study. The purpose of this study

appears to be in line with an exploratory study as well as an explanatory study. As an

operations management study, this study seems to be empirical normative quantitative

research and empirical quantitative case study.

Since a case study is suitable for “how” and “what” questions as well as exploratory research

and explanatory research, which are the cases of this study, this study chooses a case study as

17
the research strategy.

Theories relevant to this study are available from the existing theories, although the necessary

theories are not well-developed for the context of this study. However, theories from similar

contexts can be the theoretical foundation of this study. The theory of this study can be

formed by an inductive method and can generalise as analytic generalisation.

Research design is the logical sequence that links the empirical data to research questions

(Yin, 2003). The four components of research design are reviewed: research questions; the

unit of analysis; the logic linking the data to the research questions; and the criteria for

interpreting the findings. This study chose the single case research design, because the single

case design provides more opportunities for in-depth observation, and the single case design

is appropriate for the extreme case, the typical case and the longitudinal case which are the

cases of this study. Then, four tests which can evaluate the research design are reviewed.

Finally, the overall research procedure is described in this chapter.

18
References
Bacharach, S. B. (1989) 'Organisational theories: some criteria for evaluation', Academy of
Management Review, 14, (4), pp. 496 - 515.
Bertrand, J. W. M. and Fransoo, J. C. (2002) 'Operations management research
methodologies using quantitative modeling', Operations Management Research, 22,
(2), pp. 241 - 264.
Eisenhardt, K. M. (1989) 'Building theories from case study research', Academy of
Management Review, 14, (4), pp. 532 - 550.
Leonard-Barton, D. (1990) 'A dual methodology for case studies: synergistic use of a
longitudinal single site with replicated multiple sites', 1, (3), pp. 248 - 266.
McCutcheon, D. M. and Meredith, J. R. (1993) 'Conducting case study research in operations
management', 11, (3), pp. 239 - 256
Saunders, M., Lewis, P. and Thornhill, A. (2007) Research Methods for Business Students,
4th edn. Harlow, UK: Prentice Hall.
Saunders, S. (2006) Janes's Fighting Ships 2006 - 2007. Coulsdon, Surrey, UK: Jane’s
Information Group.
Slack, N., Chambers, S. and Johnson, R. (2004) Operations Management, 4th edn. Harlow,
England: Prentice Hall.
Voss, C., Tsikriktsis, N. and Frohlich, M. (2002) 'Case research in operations management',
International Journal of Operations & Production Management, 22, (2), pp. 195 -
219.
Wacker, J. G. (1998) 'A definition of theory: research guidelines for different theory-building
research methods in operations management', Journal of Operations Management, 16,
(4), pp. 361 - 385.
Waller, D. L. (2003) Operations Management: a Supply Chain Approach, 2nd edn. London,
England: Thompson.
Yin, R. K. (2003) Case Study Research, 3rd edn. London, England: Sage Publications.

19